refactor: share execution prep for tool-call phase

Keep the current tool-call bridge contract stable while extracting shared
execution setup and tightening Anthropic forwarding regressions.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

app/anthropic_schema.py

@@ -52,10 +52,9 @@ class AnthropicMessagesRequest(BaseModel):
     stop_sequences: list[str] | None = None
     # metadata.user_id is the official hint for per-user routing / abuse tracking.
     metadata: dict[str, Any] | None = None
-    # Tools / tool_choice are accepted but we can't forward them to Lingma yet —
-    # they're preserved here so the request doesn't 422, and the flattener
-    # surfaces any tool_use blocks as `[tool_use] {...}` text so the assistant
-    # still sees the context.
+    # Tools / tool_choice are accepted for compatibility and, when forwarding is
+    # enabled, are passed upstream as tool_config; tool_use / tool_result blocks
+    # are still flattened into text so the assistant can see prior tool context.
     tools: list[dict[str, Any]] | None = None
     tool_choice: dict[str, Any] | None = None